Validating a Pipeline for Quantitative Analysis of Scalp EEG in the Genetic Epilepsies
Abstract number :
2.351
Submission category :
12. Genetics / 12A. Human Studies
Year :
2023
Submission ID :
502
Source :
www.aesnet.org
Presentation date :
12/3/2023 12:00:00 AM
Published date :
Authors :
Presenting Author: Peter Galer, MSc – University of Pennsylvania
Jillian McKee, MD, PhD – Resident And Fellow Training, Children's Hospital of Philadelphia; William Ojemann, BS – Department of Bioengineering – University of Pennsylvania; Sarah Ruggiero, MS, LCGC – Children's Hospital of Philadelphia; Saurabh Mallela, Undergraduate – Undergraduate, Department of Bioengineering, University of Pennsylvania; Shiva Ganesan, MS – Children's Hospital of Philadelphia; Michael Kaufman, MS – Children's Hospital of Philadelphia; Alexander Gonzalez, MS, MBA – Children's Hospital of Philadelphia; Brian Litt, MD – Hospital of the University of Pennsylvania; Ingo Helbig, MD – Children's Hospital of Philadelphia; Erin Conrad, MD – Hospital of the University of Pennsylvania
Rationale:
Scalp EEG is a commonly performed clinical procedure for individuals with known or suspected epilepsy. EEG review is currently performed manually which may miss important information. Recent evidence suggests that quantitative EEG biomarkers can assist in epilepsy diagnosis, predict future outcomes, and track response to therapy. This could be particularly helpful in genetic epilepsy syndromes, where clinical outcomes are heterogeneous, and we lack clear clinical biomarkers of treatment response. Leveraging a large cohort of patients with genetic epilepsy syndromes at Children’s Hospital of Philadelphia (CHOP) with well-annotated EEG, genetic, and electronic medical record (EMR) data, we developed and validated a pipeline to process and analyze EEG, extracting potential quantitative biomarkers for the genetic epilepsies.
Methods:
We examined routine clinical outpatient scalp EEG from individuals with clinically verified pathogenic variants in SCN1A (n=48 individuals; 86 EEGs), SYNGAP1 (n=11 individuals; 21 EEGs), and STXBP1 (n=15 individuals; 56 EEGs) and controls without epilepsy or intellectual disability (n=104; 108 EEGs) from CHOP. We developed a pipeline to perform signal processing and feature extraction on EEGs. To validate the accuracy of our pipeline, we automatically calculated the posterior dominant rhythm (PDR) from segments of EEGs following clinical annotations of eye closure by extracting the dominant frequency band in the posterior electrodes. We then compared these detected PDR values against those in the clinical EEG report.
Results:
We developed a pipeline to efficiently clean and extract quantitative features from clinical scalp EEG and relate this output with individuals’ genomic and EMR data. We extracted 894 marked periods of eyes closed in EEGs with clinically annotated PDRs. Comparing PDRs marked by clinicians and those extracted from our pipeline, we calculated an R2 of 0.43 (Figure 1A). We found that on average, the automatically-detected PDR is 0.13 Hz (SD=1.85 Hz) less than that annotated by clinicians (Figure 1B).<
Genetics